Introduction
The United States (U.S.) has a reported infective endocarditis (IE) incidence of 15/100 000 which is considered the highest global IE rate [Reference Jamil, Sultan and Gleason1]. Hospitalization rates for IE in the U.S. continue to increase, placing a significant financial burden on the healthcare system. Factors contributing to the escalating hospital charges for IE include the pre-disposing factor of injection drug use and intensive medical management requiring diagnostic tests, extended hospitalizations, long-term intensive antibiotic therapy, and surgical interventions.
The financial burden reflected in the total healthcare cost of IE-related hospitalizations has increased significantly from $1.58 billion in 2003 to $2.34 billion in 2016, with hospitalizations increasing from 34 488 to 54 405 per 100 000 during this period [Reference Alkhouli, Alqahtani and Alhajji2]. A significant factor driving the high cost of hospitalization for IE is the intensive management of the disease through antibiotics and/or surgery. While oral antibiotics can be effective, standard antibiotics can be compromised for some patients due to their inaccessibility of cardiac structures [Reference Weymann, Borst and Popov3]. In these cases, surgery is the recommended course of action. Surgical treatment is a significant factor contributing to overall hospital charges for patients with IE. The total charges for hospitalizations involving surgery for 22 825 IE patients in North Carolina between 2007 and 2017 exceeded $400 million [Reference Schranz, Fleischauer and Chu4]. Specifically, the median charges for drug use associated-IE hospitalizations requiring surgery were significantly higher at $250 994 compared to $198 764 for non-drug use associated--IE surgery hospitalizations [Reference Schranz, Fleischauer and Chu4]. The high cost for IE may also be attributed to extended length of stay in the hospital to allow healthcare professionals to monitor and manage any post-operative complications, rehabilitation, and follow-up treatment [Reference Schranz, Fleischauer and Chu4–Reference Spelman and Sexton6].
Another important factor changing the epidemiology of IE in the past decade has been injection drug use, frequently requiring higher resource utilization and multidisciplinary care [Reference Bhandari, Alexander and Annie7] resulting in substantially higher healthcare costs [Reference Gray, Rogawski McQuade and Scheld8]. Results from a study using a nationally representative sample of U.S. patients revealed that the total inpatient hospital charges for infections related to opioid abuse, including IE, increased over three-folds from $190 million in 2002 to $700 million in 2012 [Reference Ronan and Herzig9]. Statewide and single-centre studies demonstrate hospitalization costs for patients with drug use associated-IE were twice as high in comparison to non-drug use associated-IE patients. For example, the median hospital charges for drug use associated-IE hospitalizations were $47 899 compared to $26 460 for non-drug use associated-IE hospitalizations in a single- centre study conducted during 2000–2016 [Reference Gray, Rogawski McQuade and Scheld8]. Another statewide study also showed this contrast with median charges for drug use associated--IE hospitalizations at $60 333 vs. $34 968 for non-drug use associated-IE hospitalizations between 2007 and 2017 [Reference Schranz, Fleischauer and Chu4].
The higher readmissions of IE patients have been another significant source of elevated hospital costs. In a study utilizing the 2017 national readmissions database, 23% of all 56 357 hospitalizations for IE were readmitted within 30 days after their initial discharge [Reference Zaman, Amritphale and Malozzi10]. The median hospitalization cost for patients during their initial admission was $20 241 [Reference Zaman, Amritphale and Malozzi10]. However, when IE patients were readmitted, the median hospitalization cost increased to $22 059, indicating that the financial burden was much higher on readmission than initial admission [Reference Zaman, Amritphale and Malozzi10].
Additionally, living in rural areas has been shown to be a predictor of poorer overall health outcomes and higher medical costs [Reference Nenninger, Carwile and Ahrens11–Reference Bates, Annie and Jha12]. Admissions for IE in West Virginia (WV) experienced a significant increase of 456% between 2014 and 2018 [Reference Bhandari, Alexander and Annie7]. A single- centre study in southern WV recorded 462 cases of IE between 2008 and 2015, resulting in hospital charges exceeding $17 million [Reference Bates, Annie and Jha12]. On average, a total of $70 000 was spent per patient with IDU-IE but only 22% ($3 829 701) were collected [Reference Bates, Annie and Jha12]. Another WV study examining IE patients who underwent tricuspid valve operations from 2012 to 2018 at West Virginia University Medical Centre showed that drug use associated-IE patients incurred significantly higher charges (totalling $291 037) compared to non-drug use associated-IE patients (totalling $235 620) [Reference Cook, Rankin and Roberts13]. Additionally, procedures associated with drug use had lower median reimbursement, at $42 063, compared to those not associated with drug use, at $70 393 [Reference Cook, Rankin and Roberts13].
The WV healthcare system continues to sustain losses of over $8 billion in gross domestic product annually due to the ongoing opioid epidemic [Reference Wood, Mells and Dotson14, Reference Brill and Ganz15]. The high cost of hospitalization highlights the significant financial burden placed on patients and healthcare systems. The purpose of this study is to examine the healthcare burden measured as hospital charges and associated factors for patients with IE in the largest hospital in WV.
Methods
Study design, data sources, and variables
A retrospective review was conducted manually of electronic medical records (EMR) for all patients aged 18 to <90 years, who had their index admission for IE at a major university-affiliated hospital in WV between 1 January 2014 and 31 December 2018. Patients with IE were initially identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes and confirmed by detailed chart reviews. The data were abstracted for each patient from several sources within the EMR, such as laboratory tests, imaging results, note-taking on patient history, physical examination, consultations, surgeries, hospital stays, and discharge summaries. The number of subsequent admissions was recorded to obtain data on the frequency of readmissions during the study period. Information collected from EMR was entered into a HIPAA-compliant, secure dataset using the Research Electronic Data Capture (REDCap) system [Reference Harris, Taylor and Thielke16].
The outcome variable was hospital charges defined as total charges billed to the patient account for their IE hospitalization corresponding to their admission and discharge dates in the EMR. Hospital charges were explained by characteristics related to the year of admission, demographics, substance use, clinical features, and hospital utilization. Demographic characteristics were classified by sex (male and female) and age (18–44; 45–64; >64 years). Substance use was described by smoking status (current smoker, ex-smoker, and non-smoker), alcohol use (current alcohol use, prior alcohol use, and no alcohol use), history of drug use (yes/no), and type of drug used. Clinical characteristics were explained by comorbidities, number of comorbidities, psychiatric disorders, causative organism, and type of IE (tricuspid, mitral, aortic, and pulmonic). Lastly, hospital utilization was broken down by consultations received; surgery for IE (yes/no), length of hospital stay, length of ICU stay, discharge status (alive, against medical advice, and death), and readmission during the study period.
Statistical analyses
Descriptive statistics was performed on categorical variables as counts and percentages. Continuous variables were presented as median and interquartile range (IQR). The primary outcome variable, hospital charges, was a continuous variable, presented separately as total charge as well as average (median) charge per person and the interquartile range for each variable. The medians were compared using the nonparametric independent samples median test. Variables with significant associations were selected in the multiple regression model. Multiple linear regression analyses were conducted to examine the association between the key dependent variable, hospital charges, and key predictor variables/potential confounders: age (continuous), history of drug use (yes/no), length of stay (continuous), surgery (yes/no), and number of consultations (0–5, 6–8, ≥9).
Assumptions meeting multiple linear regression were tested. These tests included sample size appropriateness, linearity, multicollinearity, normality of residuals, and testing for regression outliers. The initial study sample had 492 patients. Regression outliers were tested using leverage plots, studentized residuals, and Cook’s D. The five lowest observations and one highest observation in the residuals were identified as outliers and excluded from the analyses. Therefore, this study had 486 patients. The data normality of residuals was tested using the P–P plot. Natural log transformation was applied to the variable ‘total hospital charges’. The relationship between the length of hospital stay and hospital charges was curvilinear. Hence, to improve the linearity between the dependent and independent variables, and to boost the validity of statistical analyses, natural log transformation was also applied to the variable length of hospital stay. The multicollinearity assumption was tested using the Variance Inflation Factor. The predictor variables of interest were less than 5. All assumptions to use a multiple linear regression analysis were met in the model.
The final model included the dependent variable, total hospital charges, with the following variables that were significant in the bivariate association (age, alcohol use, length of hospital stay, surgery, substance abuse, number of drugs, and psychiatric disorders). Omnibus model fit was addressed with regression F test and an alpha of ≤0.05 was utilized. Adjusted R Square was used to test the effect size of the overall final model. Unstandardized and standardized beta coefficients with standard error were calculated for the multiple regression analyses. The relative effect of each factor on hospital charges was assessed by calculating: (a) the percent change for categorical variables compared to the reference category; or (b) per x units of a continuous variable (e.g. age in years) or log-transformed units of hospital length of stay. The percent change was calculated using the following formula: Percent Change = (exp (beta-k × units-k) − 1) × 100 where beta-k is the regression coefficient estimate and units-k are the units of measure for the regression coefficient k. The percent change for categorical variables was evaluated relative to the reference category and was interpreted alongside the intercept reference. Statistical significance was considered when the 95% confidence interval (CI) did not contain 0. All statistical analyses were conducted using SPSS software, version 28.0.
Results
The data were analyzed from 486 patients with IE who were admitted between, 1 January 2014 and 31 December 2018. The overall sum of total charges for all participants was $112 million (Median $198 678, IQR $213 309). Figure 1 represents the trend of total charges over the years, revealing a consistent increase in IE expenses over time. Baseline patient characteristics are presented in Table 1. From 2014 to 2018, IE admissions increased from 28 to 179. Healthcare charges rose from $3.7 million in 2014 to $48 million in 2018, although inflation only increased by 1.1% (0.8% in 2014–1.9% in 2018). While the number of patients increased 5.4 times, the associated hospital charges increased 12-fold over the study period.
Total hospital charges (in Million U.S. Dollars).
Baseline characteristics
Note: Charges for the variable ‘Type of Drug’ exceed 100% because drug categories are not mutually exclusive.
a Interquartile Range.
b Values in bold indicate statistical significance.
c Opiates include Fentanyl, Carfentanyl, Heroin, Codeine, Morphine, and Oxycodone.
There was a similar distribution of male and female patients with IE. Participants aged 18–44 years comprised almost two-thirds (62%) of the total patients and hospital charges. Additionally, over two-thirds (66%) of patients were current smokers, representing a majority of total charges among this subgroup. An even higher proportion (70%) of patients had a history of drug use, and they incurred over three-quarters of the hospital charges ($85.5 million). Table 1 also presents the types of drugs used by participants prior to their hospital admission. Among these patients, 59% had used opiates, including fentanyl, carfentanyl, heroin, codeine, morphine, and oxycodone, accounting for healthcare charges totalling $74 million.
The clinical and surgical characteristics of the participants are presented in Table 2. Comorbidities played a significant role in the financial burdens associated with IE. The presence of multiple comorbidities in individuals corresponded to a greater increase in overall charges. Over 58% of the patients had some form of psychiatric disorder, which incurred hospital charges amounting to $70 million. Of the patients with psychiatric disorders, 44.6% had substance use problems. About 42% of the patients had tricuspid valve-affected IE incurring the most charges, followed by mitral valve-affected IE (34%). Among the causative organisms, methicillin-resistant Staphylococcus aureus (MRSA) associated IE incurred the highest charges, followed by methicillin sensitive S. aureus (MSSA)-associated IE.
Clinical characteristics
Note: Total Charges do not add to 100% for variables where categories are not mutually exclusive.
a Interquartile range.
b Values in bold indicate statistical significance.
Hospital utilization and associated charges are presented in Table 3. Infectious disease consultations incurred the highest hospital charges, followed by cardiac surgery, social work, physical and occupational therapy, and cardiology and psychiatry consultations. About 47% of the patients underwent surgery incurring 70% of the total charges. The longer length of hospital stays caused a greater increase in overall total charges. Patients with hospital stays of 50 days or longer accounted for a third of all charges totalling $37 million. Over one-third of the patients (36.6%) were hospitalized for 40 or more days. The majority of the patients (79%) were discharged alive, accounting for charges of $88 million.
Hospital utilization characteristics
Note: Total Charges do not add to 100% for variables where categories are not mutually exclusive.
a Interquartile range.
b Values in bold indicate statistical significance.
The multiple linear regression model accounted for 85% of the linear variance in the dependent variable (natural log of total hospital charges) adjusting for model complexity (Adjusted R2 = 0.849), indicating a large association between the predictor variables selected in the model (Table 4). In the adjusted model, the percent increases in median charges for consultations were as follows:1–5 consultations, $89 865; 6–8 consultations, $211 503 (11.18% increase); and ≥ 9 consultations, $321 965 (30.87% increase). Compared to patients who were discharged home, there was a 60.32% increase in the median charges for patients who died in the hospital. Similarly, relative to patients treated by medical management alone, patients who had surgical intervention incurred an 81.85% increase in median charges (beta coefficient 0.598, 95% CI 0.528–0.669). Within the full model, a statistically significant association was observed between total hospital charges and length of hospital stay (beta coefficient 0.662, 95% CI 0.617–0.708).
Multiple linear regression of hospital charges
Note: Dependent Variable: Hospital charges is natural log transformed. Adjusted R Square: 0.849. B#: Unstandardized Beta Coefficients. β##: Standardized Beta Coefficients.
a Values in bold indicate statistical significance.
Discussion
Results from this study provide valuable insights into the evolving landscape of healthcare charges amounting to $112 million associated with IE admissions over a 5-year period. In this study, the admissions at one hospital system with a catch basin of the state of WV rose from 1.56 admissions/100 000 population in 2014 to 9.94/100 000 in 2018. Additionally, the median charge per IE patient in the WV hospital system grew from $81 965 per patient in 2014 to triple that amount ($244,632) in 2018. With a minor rise in inflation during that period (0.8% in 2014 and 1.9% in 2018), this is an extreme rise in patient management charges. The hospital charges in West Virginia for patients with IE are substantially higher than the national estimates. At the national level, the mean inflation-adjusted cost per admission decreased from $45 810 in 2003 to $43 020 in 2016 [Reference Alkhouli, Alqahtani and Alhajji2].
The demographic and clinical characteristics of IE patients shed light on the factors driving these escalating costs. Surgery significantly contributed to the rising financial challenges associated with IE in our study. Nationally, IE surgeries increased by 1.7-fold during 2011–2018 [Reference Geirsson, Schranz and Jawitz17]; however, there was a 26-fold increase in WV during 2014–2018 [Reference Bhandari, Abdulhay and Alexander18]. Patients undergoing surgical treatment for IE require pre-operative and post-operative hospital care [Reference Bhandari, Abdulhay and Alexander18, Reference Grubitzsch, Christ and Melzer19], which adds significantly to the overall charges. Furthermore, patients admitted to the ICU following surgery further escalate these costs [Reference Grubitzsch, Schäfer and Claus20]. Patients undergoing surgical treatment for IE require an increased demand for hospital resources, which also contributes to the overall cost [Reference Bhandari, Abdulhay and Alexander18, Reference Grubitzsch, Schäfer and Claus20]. In a study conducted among 369 patients who underwent surgical intervention for IE, the financial burden on the healthcare system was the median charge of $60 072 per patient [Reference Kemp, Arnaoutakis and George21]. Patients who undergo surgical procedures during hospital admissions typically experience longer hospital stays compared to those who do not have surgery [Reference Schranz, Fleischauer and Chu4, Reference Gray, Rogawski McQuade and Scheld8, Reference Bhandari, Abdulhay and Alexander18, Reference Cifarelli, McMichael and Forman22, Reference Rudasill, Sanaiha and Mardock23].
Patients who had multiple comorbidities contributed to a greater increase in overall healthcare costs. Having multiple medical conditions can lead to increased use of healthcare services and higher associated costs [Reference Lehnert, Heider and Leicht24]. Treatment for individuals with multiple comorbidities requires tailored care that may extend over a longer duration [Reference McPhail25–Reference Palladino, Tayu Lee and Ashworth27], resulting in prolonged hospital stays, and increased overall charges [Reference Kemp, Arnaoutakis and George21].
A notable proportion of IE patients had a history of drug use, representing a majority of the total healthcare costs. While nationally, the prevalence of substance use among patients hospitalized for IE was 21% during 2015–2019 [Reference Palicherla, Torres and Walters28], the prevalence was substantially higher for WV patients [Reference Bhandari, Alexander and Annie7]. Treating IE in individuals who use drugs requires extensive and costly multidisciplinary care [Reference Rodger, Glockler-Lauf and Shojaei29]. A previous study found that in 2015, patients with drug use-associated IE accounted for a total of $9.3 million in healthcare costs [Reference Fleischauer, Ruhl and Rhea30]. The cost of hospitalizations for cases of IE that are associated with opioid use doubled between 2003 and 2016 [Reference Nenninger, Carwile and Ahrens11]. The amount of money spent on hospitalizations for opioid-associated infections has increased at a much faster rate than the actual increase in the number of individuals being hospitalized for these issues [Reference Ronan and Herzig9]. Even after accounting for inflation, the rise in costs remains high relative to the increase in hospitalizations [Reference Ronan and Herzig9]. Factors such as escalating treatment expenses, greater resource use, and increased severity of illness all contribute to these elevated costs [Reference Ronan and Herzig9].
In this study, MRSA-associated IE incurred the highest charges. MRSA infections can be serious, and there are significant costs associated with treating these infections. In a study on patients hospitalized in the Department of Veterans Affairs healthcare system in the U.S., MRSA was the pathogen with the highest aggregate cost of $1.2 billion for community-onset infections and $580.2 million for hospital-onset infections in 2017 [Reference Nelson, Hatfield and Wolford31]. The authors concluded that MRSA infections had the highest total aggregate cost nationally because of their heavy financial burden.
Lastly, specific geographic and cultural aspects of the rural landscape add a burden to healthcare costs. Researchers have frequently concluded that rural populations are more likely to be poor, lack health insurance, engage in risky behaviours, and be sicker compared to their urban counterparts [Reference Lister, Weaver and Ellis32, Reference Nielsen, D’Agostino and Gregory33]. Patients in rural areas have to travel greater distances to access healthcare, which is costly and requires time away from work. In addition, rural topography and transportation challenges add further barriers to accessing healthcare centres. Rural populations also face decreased access to home health or psychiatric resources and stigma regarding drug use [Reference Gale, Janis and Coburn34]. Studies from other rural Appalachian states also indicate a similar trend in hospitalization due to IE [Reference Siddiqi, Freeman and Fanucchi35]. Since these states face more financial challenges, the burden added by the steeply rising IE hospitalizations is more significant.
Strengths and limitations
This study is strengthened by the exhaustive search of EMR from the premier hospital in WV that treats IE. Unlike most studies that have relied solely on ICD codes, our study further validated the initial IE diagnosis based on ICD codes with several clinical indicators retrieved from a comprehensive manual medical chart review of each patient. In addition, the manual chart review provided validated information on clinical and substance use-associated characteristics. Another strength of the study is that we analyzed data for individual patients and not hospitalizations. We were also able to abstract long-term data, such as the readmissions of the patients, although we did not have information on patient outcomes such as out-of-hospital mortality or subsequent admissions to other hospitals. Our study is subject to other limitations. It is a single-centre study, although the centre is the largest cardiac care centre in WV. In addition, the results are dependent on the data available in the EMR. Information was limited on socio-demographic variables, such as education, income, and duration of drug use. There was a potential for drug use data to be misclassified if patients had not disclosed previous drug use upon hospital admission. This bias was also a study limitation.
Conclusion
In conclusion, hospital charges related to IE continue to have a significant financial burden on the healthcare system. Many patients with IE in WV required complex care and experienced significant morbidity. Prolonged length of hospital stays, surgeries, readmissions, and other medical consultations and procedures resulted in high hospital charges. The authors believe that interventions need to focus on early diagnosis and treatment of patients with IE to reduce the need for complex care, complications, extended hospitalizations, and the associated high hospital charges.
Data availability statement
Data underlying this study are from an approved repository that houses clinical data from the healthcare systems in West Virginia. These data contain full Protected Health Information (PHI) and thus, legally cannot be shared publicly. Patient data are available upon request from Wes Kimble (wkimble1@hsc.wvu.edu), Director of Research Data Analytics at the West Virginia Clinical and Translational Science Institute, for researchers who meet the criteria for access to confidential data. Hospital charges data were obtained from the Finance and Reimbursement Department at West Virginia University Health System.
Acknowledgements
We thank Jennifer von Boetticher, Senior Financial Analyst at West Virginia University Health System for assisting us with the information on hospital charges for this study.
Author contribution
Bhandari R.: Study conception and design, Acquisition of data, Analysis and interpretation of data, Drafting of manuscript, and Critical revision; Abdulhay N.: Analysis and interpretation of data, and Drafting of manuscript; Smith D.: Drafting of manuscript; Wiener R. C.: Drafting of manuscript, Critical revision; and Fisher M. A.: Study conception and design, Acquisition of data, and Critical revision.
Competing interest
The authors declare no competing interests exist.
Ethical standards
West Virginia University Institutional Review Board (IRB protocol number: 1811373348) approved this study.
Funding statement
This work was supported by the National Institute of General Medical Sciences [Grant number: 5U54GM104942–08]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Open access
